HIV-1 remains a significant public health concern in Africa, with an estimated 26 million people living with HIV-1 in the region (UNAIDS, 2019). Despite the introduction of effective antiretroviral therapy (ART) and the roll-out of prevention of mother-to-child transmission programs, more than 70% of all new childhood infections occur in Africa, numbering close to 220,000 new HIV-1 infections and accounting for more than 110,000 deaths in children annually on the continent (UNAIDS, 2019). However, 1–5% of children infected with HIV-1 do not progress to AIDS for more than 10 years even without ART (Warszawski et al., 2007) – so-called long-term non-progressors (LTNPs). We have documented a cohort of childhood LTNPs from Uganda and Botswana (Kyobe et al., under review) having one of the highest documented frequencies of LTNPs (8%) in both countries. In general, LTNPs possess a high degree of HIV-1 infection control with significant viral suppression and high CD4 + T cell count (>500 cells/mL) (Warszawski et al., 2007), and have lower levels of the immune activation markers such as HLA-DR and CD38 despite chronic infection (Hua et al., 2014); this pattern is frequently associated with the expression of broadly neutralizing antibodies (bNAbs) (González et al., 2018).

HLA class I molecules predominantly function to present endogenously processed antigens to cytotoxic CD8 + T lymphocytes (CTL) in a cell-mediated immune response (Frater et al., 2007), and play a significant role in the innate immune response through interactions with natural killer (NK) cell receptors via the killer cell immunoglobulin-like receptors (KIRs) (Boudreau and Hsu, 2018). The latter interaction puts class I molecules at the epicenter of HIV-1 control and makes them a focal point for developing and designing epitope-based vaccines (Kaseke et al., 2021). As a result, genetic variation at HLA class I has been frequently associated with LTNP status. Several studies, mainly conducted among adults of European ancestry, have reported associations between LTNP and HLA-B^∗57, B^∗27, B^∗35, B^∗58, and some HLA-C alleles (Goulder and Walker, 2012). HLA class I genotypes, however, vary substantially across populations, and the association between specific alleles and HIV-1 disease progression in African populations remains controversial, with previous studies mostly conducted in adult populations, showing country/region-specificity of results utilizing small sample sizes and candidate allele approaches (Serwanga et al., 2009; Matthews et al., 2011; Peterson et al., 2013; Sampathkumar et al., 2014; Shepherd et al., 2015). For example, HLA-B^∗58:01 has been associated with slow progression in Uganda and South Africa (Serwanga et al., 2009; Payne et al., 2014), but this observation was not seen in Kenya and Botswana (Peterson et al., 2013; Payne et al., 2014); however, the initial Ugandan study was based on only 14 slow progressors, and the definition of LTNP was less stringent than comparable studies in Europeans (Serwanga et al., 2009). Notably, the HIV-1 epidemic in East Africa is predominantly due to HIV-1 subtypes A and D, while HIV-1 subtype C is predominant in southern Africa (Bbosa et al., 2019).

The now established protective effects of HLA-B^∗57 in adults are also known to be population-specific - HLA-B^∗57:01 is associated with control in Europeans while HLA-B^∗57:03 is mostly strongly associated among Africans (Matthews et al., 2011). Most importantly, none of the canonical HIV-1 protective alleles (HLA-B^∗57, B^∗27, B^∗35, B^∗58) were replicated in a cohort of 123 HIV-1 clade C infected adolescents in Zimbabwe; instead, HIV-1 LTNP was attributed mainly to the HLA-C locus (HLA-C^∗08:02 and -C^∗08:04) (Shepherd et al., 2015). Adland and colleagues provided evidence of discordance in HLA-HIV associations between children and adults (Adland et al., 2015) in a South African cohort, among whom the putative protective effects of B^∗57, HLA-B^∗58:01, and HLA-B^∗81:01 in 47 HIV-1 clade C infected mothers were non-existent in 84 children (Adland et al., 2015). HLA homozygosity has also been associated with rapid progression to AIDS, while heterozygosity has been associated with protection; however, HLA homozygosity is very low in Africa (Carrington et al., 1999; Cao et al., 2001; Peterson et al., 2013). Taken together, the current body of literature fails to provide a clear picture of the role of HLA alleles in pediatric HIV-1 LTNP, particularly among large African pediatric population groups (Thobakgale et al., 2009; Adland et al., 2015; Shepherd et al., 2015).

We have demonstrated uncaptured genetic variation in Botswana and Uganda (Retshabile et al., 2018) from exome sequencing data, which, alongside other studies of genetic variation across the continent (Sherman et al., 2019a; Choudhury et al., 2020), suggest a potential role for novel genetic variants in disease susceptibility and argue for the use of multi-ethnic African populations in such studies (Sherman et al., 2019b). We hypothesized that looking for genetic association between HLA and pediatric LTNP in multi-ethnic African populations could yield novel HLA variant associations, which might be relevant for epitope-based vaccine designs. To evaluate this, we inferred HLA class I alleles from exome sequencing data derived from a cohort of about ∼800 pediatric HIV-1 positive children from Uganda and Botswana, and determined evidence for an association between HLA class I variants and LTNP status.

A total of 1000 participants, equally drawn from both countries, were recruited to participate in this stage of the study. As summarized in Table 1, we found that gender and country were similar at the geographic level, but there were significant differences in the number of LTNPs and RPs in the two countries (p < 0.001). Consistent with clinical observations, there were more female LTNPs than male. By definition, the median time to progression among LTNPs is significantly different from RPs, and we confirmed this in our cohort (157 months vs. 17 months, p < 0.001, Wilcoxon rank–sum test). Exome sequencing was successful for 394 LTNPs (207 Uganda, 187 Botswana) and 420 RPs (162 Uganda, 258 Botswana), and this final set of 814 individuals was utilized for downstream association analyses. Eight (∼1%; 8/822) samples were excluded from further analysis because no allele was typed at any locus. A total of 68 HLA-A, 88 HLA-B, and 63 HLA-C alleles were detected in our cohort.

Using HLA typing derived from exome sequencing, we provide what to our knowledge is the largest survey of HLA allelotypes in pediatric HIV-1 infection in Africa. The sample sizes employed were more than sufficient to identify reported strong effects on disease progression. We confirmed previously known associations (HLA-B^∗57:03 and B^∗58:01) alongside a novel C^∗03:02 association with LTNP. The effect sizes of the protective alleles and haplotypes ranged from 1.89 to 5.4, and, consistent with previous studies, these variants explain up to 16.5% of LTNP variation in our cohort. Our results expand upon the known and anticipated alleles and effect sizes in comparable HIV-1 infected pediatric populations in Africa (Shepherd et al., 2015).

The replication of previously reported HLA-HIV-1 associations (HLA-B^∗57:03, B^∗58:01, and B^∗15:10) in African populations supports our methodological approach of using exome data for HLA typing in these cohorts. We observed slightly higher protective effects (OR 3.21) of HLA-B^∗57:03 in our gender-mixed cohort than expected from the female sex workers cohort in Pumwani Kenya (OR 2.68) (Peterson et al., 2013). In Southern African adults, HLA-B^∗57:03 and B^∗58:01 have been associated with low set-point viral load and a higher CD4⁺ T-cell count during early infection - two phenotypes that positively correlate with LTNP status in adults (Tang et al., 2010; Chopera et al., 2011). Indeed, functional studies among adult LTNPs and elite controllers expressing B^∗57:03 alleles have demonstrated that this allele presents epitopes that elicit superior CD8⁺ T-cell cytotoxicity activity (Migueles et al., 2015). These HLA-B associations are consistent with previous studies among adults infected with HIV-1 in Uganda, Zambia, Kenya, and South Africa (Serwanga et al., 2009; Leslie et al., 2010; Tang et al., 2010; Matthews et al., 2011; Peterson et al., 2013), as well as other Caucasian populations (Klepiela et al., 2004; Lazaryan et al., 2006; Chopera et al., 2011; Carlson et al., 2012; Gijsbers et al., 2013) and collectively, support the notion that heterozygous states of HLA protective alleles are advantageous, probably in both children and adults (Carrington et al., 1999; Arora et al., 2020), likely by providing alternative or diverse epitope presentation pathways especially in the emergence of HIV-1 escape mutants (Carlson et al., 2012; Gijsbers et al., 2013; Arora et al., 2020). This is the first report of HLA-B associations in pediatric HIV-1 infected populations from Africa (Adland et al., 2015; Shepherd et al., 2015).

The HLA-C^∗03:02 allele was associated with a fourfold increase in LTNP status among carriers; a similar association has not been reported in LTNP studies in adults. HLA-C^∗03:02 has been associated with a plethora of immune and immune-mediated phenomena, including methimazole-induced hepatotoxicity in patients treated for Graves disease (Li et al., 2019), lower BMI (Shen et al., 2018), and the development of eclampsia, allopurinol-induced SJS, and toxic epidermal necrolysis (Cristallo et al., 2011; van der Zwan et al., 2018). Other HLA-C alleles - C^∗08:02 and C^∗08:02 - have been reported as the main drivers of HIV-1 control in adolescents from Zimbabwe (n = 126) (Shepherd et al., 2015). In our cohort HLA-C^∗08:02 did not reach statistical significance after correction for multiple testing (Table 3). HLA-C^∗03:02 and C^∗08:02/04 molecules differ at positions 35 (R35Q), 114 (D114N), 116 (S116F), and 163 (L165T), located in pockets B, C, D, and E, respectively of the protein binding groove, suggesting that these alleles may have differing peptide binding affinities for different HIV-1 epitopes (Kloverpris et al., 2012). Higher extracellular surface expression of HLA-C molecules has been proposed and demonstrated as the most likely mechanism in HLA-C-mediated delayed HIV-1 disease progression (Thomas et al., 2009; Apps et al., 2013), however, the mean expression of HLA-C^∗03:02 is lower than C^∗08:02 (Apps et al., 2013) on CD3⁺ cells; therefore there may be additional mechanisms through which HLA-C^∗03:02 can control HIV-1 in African children, such as microRNA regulation of HLA-C expression and KIR recognition (Kulkarni et al., 2011; Li et al., 2018; Vargas et al., 2020).

HLA-C^∗03:02 was found in weak LD with B^∗58:01 (r² = 0.18), and both were associated with long-term control of HIV. Interestingly, the haplotypic effect of HLA-C^∗03:02 in combination with B^∗58:01 is similar to the individual allelic effect of C^∗03:02 (OR 4.88 vs. 4.78), and the allelic effect of carrying C^∗03:02 is twice that seen among carriers of B^∗58:01 (OR 4.68 vs. 1.72). However, our analysis of carriers of C^∗03:02 without B^∗58:01 showed that the associations were lost despite the strong association with protection observed when the allele was assessed agnostic to other alleles or as part of a haplotype. HLA-C^∗03:02 also showed an apparent additive effect when combined with other putatively protective alleles such as HLA-A^∗74:01, and this apparent synergistic effect of C^∗03:02 was consistently observed in combinations with detrimental alleles (Figure 6). These observations, coupled with the overall heterozygous advantage of the HLA-C^∗03:02 allele (p = 5.7 × 10^–4), lead us to postulate that HLA-C^∗03:02 may play a synergistic role in LTNP in our population groups.

We only observed a single susceptible allele and haplotype (enriched among rapid progressors) in our cohort - we confirm a previous independent association between HLA-B^∗15:10 and rapid progression with a 55% reduction in the odds of LTNP being observed in our cohort. This is congruent with prior data from Kenya (OR 0.45 vs. 0.49) (Peterson et al., 2013), which may partially reflect the predominant HIV-1A and D clades seen in both Uganda and Kenya (Sampathkumar et al., 2014). This association was also seen in Botswana, where the predominant clade is HIV-1C (Serwanga et al., 2009; Payne et al., 2014).; some HLA alleles are known to bind and present epitopes from multiple clades [allelic promiscuity (Chappell et al., 2015)], which could explain the broad applicability of this observation, as such alleles could have poor specificity and so potentiate viral replication (Honeyborne et al., 2006). HLA-B^∗15:10 was found to be expressed at lower levels in cells deficient of transporter associated with antigen processing 1 (TAP1) and TAP2, implying that B^∗15:10 might utilize TAP-dependent pathways for loading epitopes (Geng et al., 2018). TAP-dependent pathways are frequent targets of viral immune evasion (Verweij et al., 2015). Haplotype A^∗29:01-C^∗17:01 was also associated with rapid progression; consistent with Sampathkumar et al. who found that a C^∗17:01 containing haplotype was associated with faster HIV-1 seroconversion among sex-workers in Kenya (Sampathkumar et al., 2014).

Some limitations of our study merit mention. Given the long-term follow-up in our cohort, the study is limited by survival bias, therefore, the associations observed could be due to enrichment of protective HLA variants. HLA genotypes were inferred from exome sequence data, as opposed to other sequence-based or sequence-specific primer typing techniques. Despite the enrichment for exonic regions in the genome during exome sequencing, the high polymorphism in the HLA region makes such data less reliable, potentially leading to misclassification. Current evidence, however, suggests that this may not be as significant a concern as initially feared (Duke et al., 2016; Sverchkova et al., 2019; Tokić et al., 2020), and we observed a high degree of concordance between inferred and allelotyped alleles in our validation experiments. While HWE checks are useful for validating genotypes/genotyping in disease-free controls, we did not have access to suitably matched healthy cohorts and deviations from HWE are not always sufficiently sensitive to case ascertainment status, particularly with dense datasets. Instead we performed a limited validation of the results using an orthogonal method of allelotyping, which showed almost perfect concordance with our inferred allelotypes. Thus, we do not believe the allelotyping to be generating false positive results. Given the modest sample sizes employed, we were underpowered to stratify the haplotype analyses by country; independent replication of our findings in other larger pediatric populations is thus highly desired. Unfortunately, WHO changes on ART initiation (test and treat strategy) (Kretzschmar et al., 2013) mean that such studies will necessarily need to rely upon large retrospective datasets, which are relatively scarce within this age group.

The primary objective of our study was to evaluate enrichment of class I HLA loci on HIV disease progression (rather than HIV acquisition); necessarily, this required all participants to have the exposure of being HIV positive. A better context for our results would be comparison of allele frequencies among unaffected (normal) individuals from our population groups; however, there are no unaffected pediatric cohorts from our cohort countries, and comparisons across generations (adults vs. children), particularly in the context of HIV where there is likely to be ascertainment differences between sites, are made with trepidation. None-the-less, we provide population frequencies of the highlighted alleles from healthy available populations for completeness (Supplementary Table 5). Unlike Uganda, there is no comparable HLA frequency data from Botswana in the Allele Frequency Database (AFND) (Gonzalez-Galarza et al., 2020). We therefore chose to use South African Blacks or Zimbabwe Shona from the AFND and adult data from Uganda and Kenya (Luo and Nandi) as proxy populations for our Botswana and Uganda cohorts, respectively. In Botswana, this is undoubtedly imperfect, although the black South African populations are believed to be closely related to the Tswana people in Botswana, with some evidence of linguistic links across countries (Table 4 and Supplementary Table 5).

Overall, the protective alleles and haplotypes reported here account for <20% of LTNP in our cohort, leaving more than two-thirds of LTNP unexplained. Therefore, we believe that as yet unrecognized host and other factors could play key roles in determining HIV-1 progression in African children. For instance, we have recently documented an abundance of Anelloviridae viral species among LTNPs in our cohort (Mwesigwa et al., 2021). Anelloviridae are thought to impact NK cell activity through NF-kB, and, since HLA is known to interact with killer immunoglobulin-like receptors (KIR) on natural killer (NK) cells, the synergy between these two mechanisms could provide an alternative pathway for clearance of virally infected cells via NK cell cytotoxicity (Paximadis et al., 2011). Exploring these and other factors in our population cohorts may yield additional factors that mediate pediatric LTNP status in Africa and provide a potential path to new vaccines and therapeutics.

We provide evidence for the benefit of geographically independent multi-ethnic African populations to unravel novel HLA and HIV-1 disease associations. We confirm known and identify novel HLA associations with LTNP in two African pediatric populations. Consistent with emerging evidence, we demonstrate the role of HLA-C in the control of HIV-1 infection in children. Our results bolster a growing body of literature in support of an important role for HLA-C alleles in the control of HIV-1 among children and adolescents.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://h3africa.org/wp-content/uploads/2018/05/App-D-H3Africa-Data-and-Biospecimen-Access-Committee-Guidelines-final-10-July-2017.pdf, NA.x.

The studies involving human participants were reviewed and approved by Makerere University School of Biomedical Sciences IRB, Uganda National Council for Science and Technology (UNCST), University of Botswana IRB, Botswana Health Research and Development Committee (HRDC), and Baylor College of Medicine IRB. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

SK, MW, MJ, GK, AK, BN, MM, GA, IK, SWM, MT-J, CB, GM, and NH: conceptualization, writing – review and editing, funding acquisition, investigation, and project administration. SK, SM, ErK, MW, FK, AM, BN, JF, GR, BM, LW, and KM: data curation. SK, SM, and NH: formal analysis. JK-L: funding acquisition. EW and JK-L: investigation. SK, GK, GR, BN, MM, CB, GM, and NH: methodology. EW: project administration. GM, NH, MJ, AK, MM, and JK-L: supervision. SK and NH: validation and writing – original draft. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.